As cited
Copy frozen at (site build).
ai security
The AI code vulnerabilities that grow with your app
Theori tested five AI coding agents from Anthropic and OpenAI by having them build 28 applications across different scenarios, then performed penetration testing on the results. The study found that common injection vulnerabilities like SQL injection and cross-site scripting were rare, as the models typically used prepared statements and object-relational mapping frameworks.
Why it matters: Development teams evaluating AI coding assistants need to understand their actual security characteristics and residual vulnerabilities beyond traditional injection flaws, which may persist in production deployments.
- Source published
- First seen by Cybersecurity Tracker
Source attribution
Correction
Correction recorded as of .
ai security
The AI code vulnerabilities that grow with your app
No summary had been written when this copy was frozen.
- First seen by Cybersecurity Tracker
Source attribution
Correction
Correction recorded as of .
ai security
The AI code vulnerabilities that grow with your app
Theori built 28 applications using artificial intelligence (AI) coding agents from Anthropic and OpenAI, then scanned them for security flaws across different development scenarios. The AI models largely avoided common injection vulnerabilities like SQL injection and cross-site scripting, instead favoring prepared statements and object-relational mapping (ORM) frameworks. The findings suggest AI-assisted code generation may produce fewer classic web vulnerabilities than expected, though the research did not comprehensively assess other attack surfaces.
Why it matters: Development teams relying on AI coding assistants need to understand their actual security properties; this research indicates these tools avoid some traditional classes of bugs but does not establish whether they introduce new vulnerability classes or systemic risks.
- Source published
- First seen by Cybersecurity Tracker